Clinical Evaluation of Multiple Inflammation Biomarkers for Diagnosis and Prognosis for Patients with Systemic Inflammatory Response Syndrome

Abstract

A multiplexed biomarker bundle consisting of nine different inflammation markers was evaluated regarding their diagnostic and prognostic performances in 159 adult systemic inflammatory response syndrome (SIRS) patients enrolled at the emergency department. Fibronectin, interleukin-8 (IL-8), biotin, and neutrophil gelatinase-associated lipocalin (NGAL) were the most robust markers but were not superior to the already established markers IL-6, C-reactive protein (CRP), procalcitonin (PCT), and soluble urokinase plasminogen activator receptor (suPAR).

TEXT

Bloodstream infection (BSI) in patients with systemic inflammatory response syndrome (SIRS) is associated with substantial morbidity and mortality (1). Early and reliable diagnosis for appropriate treatment is essential for improving the prognosis of infected patients. While so far, no single biomarker was able to replace time-consuming blood cultures, some, like procalcitonin (PCT), C-reactive protein (CRP), and interleukin-6 (IL-6), are commonly used for additional assessments (2, 3).

Recently, a new CE-marked in vitro diagnostic (IVD)-approved multiplex analysis system (Anagnostics Bioanalysis GmbH, Vienna, Austria) for testing inflammatory biomarkers was introduced to the market, using competitive and sandwich immunoassays. This test system is a fully automated instrument that may provide extended and rapid inflammation monitoring in a single run; the time to result is 3 min per sample, and 8 samples can be measured consecutively in one run.

The objective of this study was to evaluate the diagnostic (for positive blood cultures caused by Staphylococcus aureus, Escherichia coli, or Candida spp.) and prognostic performances of the biomarker bundle in patients fulfilling the SIRS criteria (all patients had fever and fulfilled one or more additional SIRS criteria) at the emergency department (ED).

(This paper was presented in part at the 54th Interscience Conference on Antimicrobial Agents and Chemotherapy [ICAAC], 5 to 9 September 2014, Washington, DC [14].)

The biomarkers were evaluated in 159 adult patients who were included consecutively at the Medical University of Graz, Austria, between June 2012 and March 2013. All the biomarker levels (routinely tested and multiplexed) and blood culture results originated from blood samples taken when patients presented to the emergency department, before anti-infective treatment was initiated. The additional inclusion criteria were (i) no anti-infective treatment during the 5 days before blood samples were obtained, (ii) informed consent (approved by the ethics committee, Medical University Graz, Austria, no. 21-469 ex 09/10), (iii) positive blood culture results with growth of S. aureus, E. coli, or Candida spp., or negative blood culture results, and (iv) the anticipated limit of about 50 patients in each study group had not been reached. Inclusion criteria (iii) and (iv) were applied only for retrospective testing of the biomarker bundle (see details below).

Three pairs of blood culture bottles (Bactec Plus Aerobic/F and Anaerobic/F) were drawn from each patient and incubated for up to 7 days in the continuously monitored blood culture instrument BD Bactec FX (BD Diagnostics, Sparks, MD). The positive blood cultures were further processed; microbiological identifications of the responsible pathogens were performed by routine laboratory measures in the microbiological laboratory, Medical University of Graz. The routine inflammation markers IL-6, CRP, PCT, and suPAR were tested immediately after blood sample collection (Cobas 8000 system; Roche Diagnostics, Rotkreuz, Switzerland, and suPARnostic enzyme-linked immunosorbent assay [ELISA] kit; ViroGates, Copenhagen, Denmark).

The plasma samples were frozen at minus 80°C for retrospective testing (performed between May and June 2013) with the multiplexed biomarker bundle at the Clinical Institute of Medical and Chemical Laboratory Diagnostics, Medical University of Graz, Graz, Austria. The bundle of biomarkers included interleukin-8 (IL-8), serum amyloid A (SAA), cystatin C, neutrophil gelatinase-associated lipocalin (NGAL), interleukin-10 (IL-10), fibronectin, thrombomodulin, biotin, and substance P. As (i) the study focused on the diagnostic and prognostic potential in SIRS patients with bloodstream infections caused by S. aureus, E. coli, or Candida spp. and (ii) the available number of immunoassays was limited, the anticipated numbers of patients included in the study was 200 (50 patients for each causative pathogen and 50 patients each with negative blood cultures, not fulfilling sepsis criteria, and serving as controls). The anticipated numbers were almost reached for all groups except Candida spp. (due to the rareness of isolation during the study period).

The plasma biomarker levels were analyzed statistically according to the patient groups of different causative pathogens using Mann-Whitney U test and receiver operating characteristics (ROC) curve analysis to reveal the area under the curve (AUC) values and their 95% confidence intervals (CI). The optimal cutoff values for the sensitivities and specificities of the biomarkers were determined with Youden's index. Univariate and multivariable logistic regression analyses were conducted for the biomarkers, and odds ratios (OR) and 95% CIs were displayed. Explanatory variables with a P value of <0.20 in the univariate logistic regression analyses were included in the multivariable models. The variables in the final logistic regression model were selected with a forward stepwise procedure.

Statistical analyses regarding the prediction of BSI in 159 patients with SIRS revealed the highest AUC values for PCT (0.694), suPAR (0.660), biotin (0.646), and IL-8 (0.625), and the same four markers also performed best in a prediction of 30-day mortality (n = 27 [17%]). Detailed results (including cutoff values) are depicted in Tables 1 and 2. In a search for the most promising combinations of biomarkers for diagnosing BSI, we multiplied the biomarkers with the highest AUCs and those with high specificity with those with high sensitivity. A combination of CRP with PCT (AUC, 0.661; 95% CI, 0.573 to 0.749), followed by PCT, suPAR, and biotin (AUC, 0.657; 95% CI, 0.567 to 0.747) resulted in the highest AUCs.

TABLE 1.

Demographic data of study population and levels of all 14 investigated biomarkers

Characteristic/biomarker levels	Pathogen/infection (group no.)					P value comparing groups 0 and 4 (if significant)
	SIRS without BSI (0)	S. aureus (1)	E. coli (2)	Candida spp. (3)	Overall BSIs (4)
No. of patients	49	49	51	10	110
Age (mean ± SD) (yr)	65 ± 18.1	64.5 ± 13.4	67.7 ± 13.9	63.3 ± 5.7	65.9 ± 13.2
No. of females/no. of males	19/30	22/27	31/20	6/4	59/51
No. who died within 30 days	3	11	10	3	24
No. of days hospitalized (mean ± SD)	11,9 ± 12	21,8 ± 17,4	12,5 ± 12,9	18,5 ± 13,3	17,3 ± 16,6
No. (%) with:
Impaired renal function (GFR < 50 ml/min)	18 (37)	31 (63)	31 (61)	4	66 (60)
Neutropenia (<0.05 × 109)	1 (2)	0	5 (10)	0	5 (5)
Malignancy	7 (14)	15 (31)	8 (16)	5	28 (25)
Liver disease	2 (4)	3 (6)	3 (6)	4	10 (9)
Cardiovascular disease	26 (53)	30 (61)	32 (63)	4	66 (60)
Hematological disease	6 (12)	4 (8)	6 (12)	1	11 (10)
Biomarker levels (median [IQR])
CRP (mg/liter)	67 (29–231)	134 (52–256)	144 (54–252)	85 (32–183)	134 (52–252)
PCT (ng/ml)	0.34 (0.12–1.18)	1.40 (0.29–7.93)	1.84 (0.42–11.30)	0.93 (0.11–1.95)	1.53 (0.32–9.12)	<0.001
IL-6 (pg/ml)	153 (81–279)	202 (110–448)	360 (113–1,320)	143 (60–365)	216 (110–851)
suPAR (ng/ml)	5.83 (4.63–7.04)	8.84 (5.67–13.99)	7.23 (4.30–11.30)	9.94 (9.43–12.01)	8.36 (5.06–12.49)	0.001
Creatinine (mg/dl)	1.15 (0.97–1.77)	1.63 (1.08–3.75)	1.29 (1.01–2.07)	0.95 (0.62–1.93)	1.50 (0.99–2.61)
IL-8 (pg/ml)	0 (0–499)	0 (0–652)	597 (379–794)	357 (0–1,050)	451 (0–756)	0.009
SAA (μg/ml)	700 (174–700)	700 (320–700)	700 (245–700)	232 (59–700)	700 (222–700)
Cystatin C (μg/ml)	1.69 (1.33–2.63)	2.26 (1.58–4.14)	1.93 (1.40–2.67)	1.64 (1.05–2.69)	2.06 (1.52–3.13)
NGAL (ng/ml)	17 (9–42)	37 (10–89)	23 (9–78)	29 (8–56)	28 (10–83)	0.046
IL-10 (ng/ml)	0.42 (0.09–2.19)	0.53 (0.12–1.73)	0.41 (0.16–1.77)	0.41 (0.05–1.11)	0.42 (0.12–1.73)
Fibronectin (μg/ml)	409 (352–444)	396 (355–436)	359 (306–420)	302 (204–343)	373 (311–433)	0.020
Thrombomodulin (ng/ml)	0.00 (0–0)	0.00 (0–0.85)	0.00 (0–1.15)	0.00 (0–0)	0.00 (0–0.85)
Biotin (ng/liter)	21 (3–51)	43 (10–333)	93 (11–261)	187 (20–602)	71 (11–281)	0.004
Substance P (ng/ml)	0.30 (0.24–0.44)	0.31 (0.23–0.44)	0.38 (0.24–0.51)	0.33 (0.3–0.38)	0.33 (0.23–0.47)

TABLE 2.

Accuracy of biomarkers in predicting positive blood cultures (diagnostic potential) and 30-day mortality (prognostic potential) in patients with SIRS, together with results of univariate and multivariable logistic regression analyses regarding bloodstream infection

Biomarker	Diagnostic potential (bacteremia/fungemia)									Prognostic potential (30-day mortality) (AUC [95 % CI])
	AUC	95% CI	Cutoff (Youden index)	Sensitivity (%)	Specificity (%)	Univariate analysis		Multivariate analysis
						OR (95% CI)a	P	OR (95% CI)	P
PCT	0.694	0.607–0.781	0.6 (ng/ml)	62	67	0.997 (0.985–1.009)	0.650		NSb	0.630 (0.516–0.744)
suPAR	0.660	0.575–0.745	7.6 (ng/ml)	55	86	1.099 (0.955–1.265)	0.188	1.149 (1.046–1.262)	0.004	0.676 (0.572–0.781)
IL-6	0.597	0.593–0.692	303.1 (pg/ml)	43	81	1.000 (1.000–1.000)	0.505		NS	0.506 (0.376–0.636)
CRP	0.586	0.488–0.683	51.6 (mg/liter)	75	47	1.000 (0.995–1.005)	0.949		NS	0.582 (0.474–0.689)
Biotin	0.646	0.550–0.743	70.4 (ng/liter)	50	81	1.000 (0.998–1.001)	0.847		NS	0.663 (0.555–0.771)
IL-8	0.625	0.534–0.716	507.2 (pg/ml)	45	77	1.002 (1.000–1.003)	0.014		NS	0.637 (0.507–0.767)
Fibronectin	0.384	0.292–0.476	377.4 (μg/ml)	54	69	0.995 (0.989–1.001)	0.096	0.994 (0.988–0.999)	0.014	0.453 (0.333–0.573)
NGAL	0.599	0.509–0.689	82 (ng/ml)	26	96	1.009 (0.997–1.022)	0.156		NS	0.605 (0.486–0.725)
Cystatin C	0.578	0.484–0.672	2.1 (μg/ml)	50	71	1.148 (0.0575–2.292)	0.696		NS	0.624 (0.513–0.735)
Thrombomodulin	0.543	0.447–0.638	0 (ng/ml)	27	81	1.021 (0.970–1.074)	0.434		NS	0.580 (0.455–0.705)
Substance P	0.524	0.426–0.622	0.3 (ng/ml)	56	53	0.680 (0.076–6.091)	0.730		NS	0.522 (0.407–0.637)
SAA	0.519	0.422–0.617	289.4 (μg/ml)	73	35	1.000 (0.998–1.002)	0.743		NS	0.524 (0.413–0.636)
IL-10	0.508	0.409–0.607	1.9 (ng/ml)	23	67	0.819 (0.606–1.106)	0.192		NS	0.532 (0.409–0.655)

^aOR, odds ratio; CI, confidence interval.

^bNS, not significant.

Only two biomarkers showed significant differences between Gram-positive and Gram-negative bacteremia, with higher IL-8 levels (P = 0.006) and lower fibronectin levels (P = 0.010) found in E. coli BSI. Fibronectin was also found to be significantly lower in the patients with fungemia than in those with bacteremia. The results of the univariate and multivariable logistic regression analyses are displayed in Table 2.

The patients who died within 30 days had significantly higher levels of suPAR (P = 0.004), biotin (P = 0.009), IL-8 (P = 0.019), PCT (P = 0.037), and cystatin C (P = 0.043) than those who survived, while no difference was found for the other biomarkers. The results of the ROC curve analysis are depicted in Table 2. Regarding the prognosis (30-day mortality), combining SAA with biotin (AUC, 0.679; 95% CI, 0.567 to 0.791) and CRP with biotin (AUC, 0.633; 95% CI, 0.513 to 0.753) revealed the highest AUCs.

With regard to the influence of renal function on biomarker levels, significant correlations with creatinine levels were found for cystatin C (Spearman's rho, 0.805), suPAR (0.586), PCT (0.522), NGAL (0.513), biotin (0.391), IL-6 (0.375), and CRP (0.27).

To summarize, out of the multiplexed bundle of inflammation markers, only fibronectin, IL-8, biotin, and NGAL showed diagnostic and, in part, prognostic potential in our investigated cohort. In particular, a decrease in fibronectin was found to be discriminatory for BSI prediction and distinction between the causative pathogens. Fibronectin plays a major role in cell growth, migration, differentiation, and wound healing. Decreased levels of plasma fibronectin are associated with acute inflammation as well as recent surgical trauma and disseminated intravascular coagulation. In this study, the fibronectin levels were highest in those without BSI, followed by those with Gram-positive and Gram-negative bacteremia, and they were lowest in those with fungemia. Similar results were found by Mamani and colleagues (4), who found lower fibronectin levels in patients with sepsis than those in patients with infectious diseases other than sepsis. Although Ruiz Martín et al. (5) discovered substantially lower fibronectin levels in septic patients (median, 102 mg/liter) than those found in this study (median, 373 μg/ml), the biomarker showed a similar trend: plasma levels were lowest in the patients with sepsis, followed by febrile patients without sepsis (median, 185 mg/liter), patients with noninfectious diseases (median, 175 mg/liter), and healthy patients (median, 256 mg/liter). These observed differences in fibronectin levels may be explained by the different detection methods used by the test systems.

The primary function of the cytokine IL-8 is to initiate chemotaxis, mainly in neutrophils, causing immunocompetent cells to migrate toward the site of inflammation/infection and induce phagocytosis. Previous studies on IL-8 investigating its diagnostic accuracy in predicting positive blood cultures in newborns suggested IL-8 as a promising asset in sepsis diagnosis (6), with high specificity (96%) but limited sensitivity (62%). In our study, specificity (77%) was not as high, and sensitivity was even lower (45%) at the optimal cutoff, which was in accordance with a larger study of critically ill patients admitted with suspected sepsis (7).

Biotin is involved in biotin-dependent transcription regulation in humans and is an essential cofactor in various microorganisms, in particular, bacteria. Biotin, for which until now only very limited data have been published, showed a sensitivity of 50% and a specificity of 81% using the 70.4-ng/liter cutoff. NGAL, a mediator of iron traffic, is expressed by immune cells, tubular cells of the kidney, and hepatocytes, and elevated levels are found during the acute-phase response. The best cutoff value for NGAL to distinguish between the positive and negative blood cultures in the SIRS patients was found to be 82 ng/ml, with a specificity of 96% but low sensitivity (26%). Mårtensson et al. (8) previously found NGAL helpful in identifying bacterial infections in critically ill patients, but they used a higher cutoff (98 ng/ml), which resulted in a higher sensitivity (77%) but lower specificity (79%).

We also found that PCT had the highest diagnostic potential of all the biomarkers evaluated. Procalcitonin (PCT) has been proposed as a useful diagnostic marker in the emergency department, although PCT may fail to rule out BSI (3, 9, 10).

Our study group recently reported suPAR as the most promising prognostic biomarker marker in early SIRS (11); consistent results were found also in this subset of patients.

Finally, combinations of new or new and established biomarkers did not show superior diagnostic or prognostic ability compared to single biomarkers. This finding is in contrast to a publication by Kofoed et al. (12) in which superiority was reported by including more than one biomarker in their AUC curves.

Some important limitations need to be addressed: (i) the fact that this is a single-center study, which attributes to the low number of patients included, and as such, this paper should be considered a preliminary report; and (ii) the biomarkers were measured in a single initial sample only but not in follow-up samples. Further testing on biomarker levels in the days following inclusion may improve the accuracy of the results (13). Other limitations include (iii) the fact that the multiplexed biomarkers were investigated retrospectively and did not have an effect on clinical decision making, (iv) a possible influence of therapeutics other than antibiotics and source of infection on the tested parameters, as well as and that correlations with severity scores were not investigated, and (v) only 3 pathogens were assessed, limiting the generalizability of the findings to other organisms causing BSI. In conclusion, out of the multiplexed bundle tested, biotin, fibronectin, NGAL, and IL-8 were the most robust markers but were not superior to the already-established markers (i.e., PCT and suPAR). The diagnostic and prognostic potentials of the investigated biomarkers were not enhanced by combining biomarkers. Larger studies may allow for a better evaluation of their clinical diagnostic and prognostic potentials in patients with SIRS.